Skip to main content
Topic: Windows 8 - Resampling, Shared Mode (Read 11138 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Windows 8 - Resampling, Shared Mode

I know this has been asked a bit, but i can´t find any definite answer to it.


So here is the thing.

As you all may know, all applications that use Sound are going through a mixer, this is since Vista i think. This is good, as it allows you to choose the volume for every application and stuff like that.
However, it´s not just dance and roses. the Mixer is it´s own stream, so it will have a bit depth and sampling rate aswell, and that means that everything will be resampled when it goes through.

So the thing i wonder is.

What should i set the settings on?

My soundcard is ZxR, not sure if it processes afterwards, i don´t think it does anything with Stereo Direct, but not absolutely sure.


But, from my tests, 16bit is a no go. It introduces noise as the artifacts are to high compared to 24bit, where it´s inaudible (at least i think so, as resampling at 16bit will often do this, that´s why you want to resample at higher bit depth to prevent this).

So then there is 24 bit.

And then to the sampling rates.

I can have from 44.1 to 192khz. And Default is 192khz for Stereo Direct.

And from my knowledge, anything over 48khz is overkill and introduces no benefit.

However, what i don´t know is if that matters for the mixer, i don´t know if it is like the bit depth, where you want to resample at higher bit depth, so maybe higher sampling rate is good as well?


And also, as resampling itself is bad, preventing it is good. So if that´s possible, should i set it to 48khz?
As that´s the highest i have in audio, most stuff are 44.1khz though. 

But maybe it resamples anyway, even if it´s the same samplerate?



Also, i am using Windows 8, from what i have read, it has improved the resampling and such compared to windows 7, though can´t speak for that myself.

Windows 8 - Resampling, Shared Mode

Reply #1
If your material is 44.1 and your device can actually run at that then use 44.1. Then again it seems to be glitchy at 16 bit so it may not support every sample rate.

Windows 8 - Resampling, Shared Mode

Reply #2
If you want to avoid resampling: set the common format to 44.1 kHz. If the track's format matches the one you configured for your soundcard, Windows Audio Engine* will not do sample rate conversion.

*) this is different from the Windows XP "kernel mixer" aka kmixer
"I hear it when I see it."

Windows 8 - Resampling, Shared Mode

Reply #3
It´s both 44.1 and 48khz, so guess i should use 48khz, but i am not sure if the soundcard resamples as well, i don´t think it does in a certain mode, but not sure, is there a way to know?

And well, don´t everyone get noise at 16bit with Windows mixer?
Try and set 16 bit, and any sample rate, you will hear a hiss. even when nothing is playing, it´s always there.

Windows 8 - Resampling, Shared Mode

Reply #4
And Default is 192khz for Stereo Direct.


No. Stereo direct is exactly that. It completely bypasses windows mixing/sampling software and does direct d/a conversion of whatever signal you are feeding it. Its necessary for pure playback of 192khz high resolution audio but does not upsample lower rates.

Windows 8 - Resampling, Shared Mode

Reply #5
don´t think it does in a certain mode, but not sure, is there a way to know?


You could try doing RMAA tests in 44.1k mode and see if the frequency response gets weird close to 22k.  If it doesn't its either not resampling, or else using such high quality resampling that the difference isn't very important. 

And well, don´t everyone get noise at 16bit with Windows mixer?
Try and set 16 bit, and any sample rate, you will hear a hiss. even when nothing is playing, it´s always there.


Thats probably just buggy hardware.

Windows 8 - Resampling, Shared Mode

Reply #6
don´t think it does in a certain mode, but not sure, is there a way to know?


You could try doing RMAA tests in 44.1k mode and see if the frequency response gets weird close to 22k.  If it doesn't its either not resampling, or else using such high quality resampling that the difference isn't very important. 

And well, don´t everyone get noise at 16bit with Windows mixer?
Try and set 16 bit, and any sample rate, you will hear a hiss. even when nothing is playing, it´s always there.


Thats probably just buggy hardware.


Could you show point me towards how to do that, i have used RMAA before i think, but i don´t really understand much of it, i think it´s the one where you get "Good" "Excellent" rating on your card?

Are you sure it´s the hardware, i have always had that for years back in different hardware.

Windows 8 - Resampling, Shared Mode

Reply #7
Could you show point me towards how to do that, i have used RMAA before i think, but i don´t really understand much of it, i think it´s the one where you get "Good" "Excellent" rating on your card?


Check their manual.

Are you sure it´s the hardware, i have always had that for years back in different hardware.


I've never seen a device where it makes a difference, so I'm inclined to think its specific to the devices you've chosen to buy.

Windows 8 - Resampling, Shared Mode

Reply #8
Could you show point me towards how to do that, i have used RMAA before i think, but i don´t really understand much of it, i think it´s the one where you get "Good" "Excellent" rating on your card?


Check their manual.

Are you sure it´s the hardware, i have always had that for years back in different hardware.


I've never seen a device where it makes a difference, so I'm inclined to think its specific to the devices you've chosen to buy.


Will do.


So if you choose 16bit in Windows now, you don´t hear any difference from 24 bit? (I am talking in idle, not playing any sound)


EDIT:

Just tried just changing to 16bit in Stereo mode (not Stereo Direct which is supposed to NOT process the audio), and that does not get any hiss.

Hmm, makes me wonder, i am pretty sure i have always had hiss with 16bit on other hardware since, at least some way back.

Maybe it can tell whether the audio is processed by the soundcard or not?

Windows 8 - Resampling, Shared Mode

Reply #9
Have to dig up this again.

The last post i said the problem only occurs with 16-bit in Stereo Direct.
This is incorrect, i thought it was just a "Stereo Direct uses 24-bit and resamples bad" and it was solved.

Now the case was simple, Stereo Direct seems to have the audio On constantly, always having the electric flow on.
Normal seem to do what soundcards usually do, they go into "sleep mode" when nothing is using the card.

So if i play something, even if the player is muted, just as long as the player says (Sound Card play this), then it will activate.
And this will bring out the same noise, meaning it's there anyway.


So, i wrote to Creative about this, will see what happens.

Never the less, the issue persists even on my onboard card.
This always have noise though, but we are talking interference, and at very low rates, this shouldn't appear on internal recordings as they are added after the processing is done, and this is normal.

I can't really think that i just happen to have 2 sound cards, 1 highend and 1 lowend and both have the same issue, doesn't it seem to much of a coincidence?

I am trying to find a solution for this, as the issue should be from something, i can't believe it's on the hardware level, it must be a driver issue or something else.


So, any suggestions?
Perhaps trying Linux or something to see if it may be a driver issue (If anyone knows linux of course, i haven't really used it).

Thanks

Windows 8 - Resampling, Shared Mode

Reply #10
Older Creative cards were quite noisy when any analog input port was left open (IIRC, not with ASIO). Does closing all input ports make any diffrence?

Windows 8 - Resampling, Shared Mode

Reply #11
Older Creative cards were quite noisy when any analog input port was left open (IIRC, not with ASIO). Does closing all input ports make any diffrence?


The card isn't old, though that depends if 2 years is old with these stuff?

Closing all input ports, you mean disable them, or disconnecting them?
Am pretty sure it shouldn't matter, as the onboard card doesn't matter with that.

ASIO for that matter, am i pretty sure only uses 24 bit with this card, and 24bit is "immune" to this issue, both onboard and ZxR.

Windows 8 - Resampling, Shared Mode

Reply #12
So does this extra noise only occur in 16 bit but not in 24 bit? That would indicate that you are hearing the 16-bit noise floor there. What are you driving with the card, sounds like very sensitive headphones / in-ears? It is not uncommon for soundcards of the better kind to have DACs with a fixed output level, and to do volume control entirely in software. The Asus Xonar ST(X) exhibits reduced dynamic range in 44.1 kHz (though still about 10 dB better than 16 bit), and some people have complained about that as well.

Unwanted noises that are not pure hiss usually originate in ground loops. Those may vary depending on the system the card is installed in - board and case in particular, including front panel.

I assume you are already running the latest bugs collection, err, drivers? (The current version for the ZxR was a Christmas present - 24 Dec 13.  The latest Realtek HDA drivers came out in April.)

If you are in fact bumping against the 16-bit noise level (and other low-level gremlins), I would suggest getting the best kind of attenuator you can buy - an amplifier. Something like the trusty O2 headphone amp, equipped with a gain selection of 1x and 2.5x. Right now your listening would seem to be taking place at more than 50 dB below DAC full scale - you could easily use 20-30 dB of attenuation. (That would be a volume pot setting of about 10 o'clock.)

Windows 8 - Resampling, Shared Mode

Reply #13
So does this extra noise only occur in 16 bit but not in 24 bit? That would indicate that you are hearing the 16-bit noise floor there. What are you driving with the card, sounds like very sensitive headphones / in-ears? It is not uncommon for soundcards of the better kind to have DACs with a fixed output level, and to do volume control entirely in software. The Asus Xonar ST(X) exhibits reduced dynamic range in 44.1 kHz (though still about 10 dB better than 16 bit), and some people have complained about that as well.

Unwanted noises that are not pure hiss usually originate in ground loops. Those may vary depending on the system the card is installed in - board and case in particular, including front panel.

I assume you are already running the latest bugs collection, err, drivers? (The current version for the ZxR was a Christmas present - 24 Dec 13.  The latest Realtek HDA drivers came out in April.)

If you are in fact bumping against the 16-bit noise level (and other low-level gremlins), I would suggest getting the best kind of attenuator you can buy - an amplifier. Something like the trusty O2 headphone amp, equipped with a gain selection of 1x and 2.5x. Right now your listening would seem to be taking place at more than 50 dB below DAC full scale - you could easily use 20-30 dB of attenuation. (That would be a volume pot setting of about 10 o'clock.)


That is correct, 24 bit makes the issue disappear.

Yes that has been my conclusion as well, but i am the only one having this problem, and that makes it just weird.

The ZxR card is going from RCA to 3.5mm into an O2 Amplifier.
The onboard one was just normal direct input into it when i tried.

I think i did some RMAA checks about this before, i don't think there was any real difference between the modes, in terms of quality, should be available here i think.

Yes i have ignored ground loop, interference and such in this, those sound completely difference, and often act differently depending on system load.
And yes, newest drivers on track.

Fun fact that you mention O2, as i already have it;P
i am currently having it at 12 o'clock, not sure what my amplifying is though, i guess it's the default, 2.5x?
This is without me pressing the Gain button.


The problem is however, that i can record this noise, while others can't.

If i record something in 16bit (The playback, not talking input), then there will be noise even in complete silence.
If others do this, there isn't, that's the thing i am bothered about.

Windows 8 - Resampling, Shared Mode

Reply #14
...
The ZxR card is going from RCA to 3.5mm into an O2 Amplifier.
The onboard one was just normal direct input into it when i tried.

I think i did some RMAA checks about this before, i don't think there was any real difference between the modes, in terms of quality, should be available here i think.
...
The problem is however, that i can record this noise, while others can't.
...


If you can hear/record that hissing sound shouldn't that be seen in RMAA results. Make another RMAA test suite and bring the results here.

Windows 8 - Resampling, Shared Mode

Reply #15
If you can hear/record that hissing sound shouldn't that be seen in RMAA results. Make another RMAA test suite and bring the results here.


I got an old test at 16-bit, doesn't tell me much, but should be the same there, as the issue as far as i know, has always been there, (Worth noting, i always had this MB as well, though i highly suspect this doesn't matter).



This image is in another topic, seems like i have 2 topics that came to the same thing in the end.
Though the two differ a bit, so shouldn't matter in the end.


Windows 8 - Resampling, Shared Mode

Reply #16
I got an old test at 16-bit, doesn't tell me much, but should be the same there, as the issue as far as i know, has always been there, (Worth noting, i always had this MB as well, though i highly suspect this doesn't matter).
...
This image is in another topic, seems like i have 2 topics that came to the same thing in the end.
Though the two differ a bit, so shouldn't matter in the end.


Hmm... when you ran that RMAA test suite, did you check that the system output format was equal with RMAA settings (i.e. no SRC involved)?

Windows 8 - Resampling, Shared Mode

Reply #17
Hmm... when you ran that RMAA test suite, did you check that the system output format was equal with RMAA settings (i.e. no SRC involved)?


You mean if i set Windows Sound Settings the same as the RMAA?
I thought it was bypassing that, exclusive use or something?

If not, then it was most likely set to 24-bit 48khz (perhaps 192khz).

Windows 8 - Resampling, Shared Mode

Reply #18
I thought it was bypassing that, exclusive use or something?

Only if you use something like ASIO or Microsoft's WASAPI in exclusive mode. (There's also the now legacy Kernel Streaming / KS).

Most if not all of the other stuff (DirectSound / DS, MME) goes through the sound engine where it is being resampled if the format does not match your settings.
"I hear it when I see it."

Windows 8 - Resampling, Shared Mode

Reply #19
I thought it was bypassing that, exclusive use or something?

Only if you use something like ASIO or Microsoft's WASAPI in exclusive mode. (There's also the now legacy Kernel Streaming / KS).

Most if not all of the other stuff (DirectSound / DS, MME) goes through the sound engine where it is being resampled if the format does not match your settings.


So it's not using the Exclusive mode then.

Hmm, does it look bad though, the test, or is it normal?
If so, then at least the resampling in windows is working correctly i guess?

And been wondering, if the Bit depth doesn't match, what happens?
for example, i have 24bit, the audio 16bit?

I read that it will simply just add empty data to make it 16bit, (unaltering the sound), is that correct?

Windows 8 - Resampling, Shared Mode

Reply #20
So it's not using the Exclusive mode then.  Hmm, does it look bad though, the test, or is it normal? If so, then at least the resampling in windows is working correctly i guess?

And been wondering, if the Bit depth doesn't match, what happens? for example, i have 24bit, the audio 16bit?  I read that it will simply just add empty data to make it 16bit, (unaltering the sound), is that correct?


  Exclusive mode is asked by software ... you need to use WASAPI to get it asked (if allowed by software).

Your RMAA results are maybe close to normal but, if there were SRC done in your test then can you say the results are truthful for other but the one mode with matching bit-depth/samplerate. Use ASIO if you re-run the test suite.

Hard to say how Windows handles bit-depth mismatch... but if you're sure that the "Stereo Direct" mode is equal with Exclusive mode then you don't have to bother this ... but is it? Zx User's Guide says:

Quote
Notes:
Stereo Direct is an option for direct playback to stereo speaker channel sampled at 192kHz.


If this is the case then is the SRC done in software or hardware?

Windows 8 - Resampling, Shared Mode

Reply #21
Exclusive mode is asked by software ... you need to use WASAPI to get it asked (if allowed by software).

Your RMAA results are maybe close to normal but, if there were SRC done in your test then can you say the results are truthful for other but the one mode with matching bit-depth/samplerate. Use ASIO if you re-run the test suite.

Hard to say how Windows handles bit-depth mismatch... but if you're sure that the "Stereo Direct" mode is equal with Exclusive mode then you don't have to bother this ... but is it? Zx User's Guide says:

Quote
Notes:
Stereo Direct is an option for direct playback to stereo speaker channel sampled at 192kHz.


If this is the case then is the SRC done in software or hardware?


Sadly, ASIO won't work in RMAA, that much i remember, i have no idea why, but i am also sure that ASIO is limited to 24bit, i think it was that, or perhaps it was 16 and 24 and 32bit didn't work:S
It did however not work in RMAA.

I actually asked that to Creative on the Forum, but didn't get an answer, though i also had an extensive discussion on ASIO with them via email, forgot to ask them that time though, will ask on the forum again (they are very hard to reach, and answers so slowly).

I Hope that it just means, it prevents resampling on Hardware level, and any manipulation, ignore all "enhancements" and just sending pure audio.
However Windows can still interfere with resampling.

If it however Always internally resamples everything to 192khz with it, then i will be disappointed and hope i can make them change it.

Windows 8 - Resampling, Shared Mode

Reply #22
Sadly, ASIO won't work in RMAA, that much i remember, i have no idea why, but i am also sure that ASIO is limited to 24bit, i think it was that, or perhaps it was 16 and 24 and 32bit didn't work:S It did however not work in RMAA.

...


Just a note: I have RMAA v. 6.3.0 free version installed and it lets select ASIO as well ... dunno if the latest version 6.4.0 still allows that.


Windows 8 - Resampling, Shared Mode

Reply #23
Yep, also in the latest version. Depending on your hard/software you may need to install the ASIO4ALL "wrapper" though: http://www.asio4all.com/

But you should be able to achieve identical results with DirectSound or even MME if you set the right sampling rate in the sound control panel and your levels are not too hot (sound engine has a limiter), but this shouldn't be a problem since the preferred level is -1 dBFS if I remember correctly.
"I hear it when I see it."

Windows 8 - Resampling, Shared Mode

Reply #24
Nowadays the free version of RMAA also supports ASIO. Should your ASIO driver prove uncooperative, you can always use ASIO4All. (I suspect that a number of "native" ASIO drivers on consumer-grade soundcards are just WDM wrappers anyway...) Note that some care is required when setting up ASIO in RMAA - be sure to pick the correct channels.

What happens in the Windows audio stack in shared mode is about the following:
Audio data from playing application is converted to float32.
Audio streams are attenuated by 3 dB [1] (to avoid clipping) and run through a mixer (i.e. added, with adjustable attenuation). Further processing may take part at this point - see APOs.
Then things are converted to the selected output format for the output device in use: Run through resampler if necessary, convert from float32 to whatever output sample format.

[1] Or at least that's what I found on Vista.

Recording is about the same backwards, except mixing + attenuation are not needed.

Therefore, make sure you choose your input and output device sample rates + formats properly when using RMAA with DSound or MME.

In exclusive mode, the application can tell the audio driver which output format to use, and the mixer and resampling are bypassed. Both the combinations of Foobar2000 w/ WASAPI and Audacity w/ASIO + ASIO4All ultimately seem to use exclusive mode.

Your RMAA results look decent enough (not having seen the graphs, which tell a lot more to the experienced eye) - but honestly you don't need a fancy card like a ZxR to get 103 dB(A) in loopback. My lowly Audigy FX ($40) will do the same - well, in 48 and 96 kHz anyway, as 44.1 recording seems to be stuck in 16 bit (and may remain like that, as the company seems to think that they only ship products with perfect drivers - there is no customer feedback path to driver development whatsoever, at least all I ever got from the friendly support drones after much discussion was an offer for a free next-model-up replacement card, which I thought nice, but who with any affinity to 44.1 kHz would want an EMU10k job like the Audigy RX?).

The ZxR uses a PCM4220 ADC, a rather expensive part specified at 123 dB SNR. Even assuming a so-so implementation, I would still expect 110-ish dB in loopback at least. Mind showing us a graph containing the noise floor (noise, DNR, THD)? Maybe it's just ground loop related garbage, though there shouldn't be too much of that with input and output on the same card. It's extremely hard to evaluate high-quality soundcards with unbalanced connections because of this - you basically don't get around test equipment with galvanically isolated ins and outs.

So you already have an O2? Good. Yes, the standard gain with gain switch out is 2.5x.
Now 12 o'clock on a typical log pot is about -20 dB, so you're getting about -12 dB overall (2.5x = +8 dB).
If your headphones are as sensitive as I think they are - you still haven't bothered to mention them - I would rather suggest a standard listening position of about 9-10 o'clock (about -30..-40 dB or so), with source volume increased accordingly.
Looks like you could consider modifying your O2 for gain selections of 1x/2.5x, assuming you are handy with a soldering iron. That should be pretty easy - at the very minimum you need to do no more than removing R19 and R23, which converts the high gain position (6.5x) to 1x. If you don't like high and low gain now being reversed, move R17 and R21 over to the R19 and R23 positions.

May I suggest that this thread be moved into Audio Hardware?

 
SimplePortal 1.0.0 RC1 © 2008-2019